09-05-2011, 12:46 PM
Abstract
In this paper, a Brain Computer Interface (BCI) isdesigned using electroencephalogram (EEG) signals where thesubjects have to think of only a single mental task. The methoduses spectral power and power difference in 4 bands: delta andtheta, beta, alpha and gamma. This could be used as analternative to the existing BCI designs that require classificationof several mental tasks. In addition, an attempt is made to showthat different subjects require different mental task forminimising the error in BCI output. In the experimental study,EEG signals were recorded from 4 subjects while they werethinking of 4 different mental tasks. Combinations of resting(baseline) state and another mental task are studied at a time foreach subject. Spectral powers in the 4 bands from 6 channels arecomputed using the energy of the Elliptic FIR filter output. Themental tasks are detected by a neural network classifier. Theresults show that classification accuracy up to 97.5% is possible,provided that the most suitable mental task is used. As anapplication, the proposed method could be used to move a cursoron the screen. If cursor movement is used with a translationscheme like Morse Code, the subjects could use the proposedBCI for constructing letters/words. This would be very useful forparalysed individuals to communicate with their externalsurroundings.
I. INTRODUCTION
Electroencephalogram (EEG) based Brain-ComputerInterface (BCI) technology has seen much development inrecent years. Specifically, EEG based BCI technologies thatdo not depend on peripheral nerves and muscles have receivedmuch attention as possible modes of communication for thedisabled [1,2,4,5,7,8,10,11]. In general, these could be dividedinto several types: mental task [1,4,7], readiness potential (murhythm) [8] and P300 evoked potential [2]. Mental task basedBCI is somewhat the least studied among the methods due tothe difficulty in obtaining low error rates. Keirn and Aunon[4] proposed a BCI design using classification of pairs ofmental tasks represented by spectral power asymmetry ratio indelta, theta, alpha and beta bands using Bayesian classifier.Anderson et al [1] studied classification of baseline andmultiplication mental tasks using neural network classificationof autoregressive features. Palaniappan et al [7] studiedclassification of three mental tasks using Fuzzy ARTMAPclassifier. Reviews of some of these technologies anddevelopments in this area are given by Vaughan et al [10] andWolpaw et al [11].In this paper, a technique is proposed using spectral powerand power difference in 4 bands from 6 channels to classifyinto two outputs, where one is baseline and another is aspecific mental task. There are some differences in the currentwork as compared to the earlier studies. The first is the use ofgamma band (30-50 Hz) EEG in addition to the commonlyused delta, theta, alpha and beta bands (below 30 Hz). Thesecond is the use of spectral power difference from 6 channelsinstead of the spectral power asymmetry (i.e. difference)between hemispheres studied by Keirn and Aunon [4]. Thethird is the study of single mental tasks instead of 2 or 3mental tasks as studied by others [1,4,7]. It is shown that it ispossible to construct a simple BCI with subjects eitherthinking of a single mental task or relaxing.Elliptic filters have been utilized to extract EEG in 4 bands:delta and theta, beta, alpha and gamma. Delta and theta bandsare combined since their frequency range is small. Thespectral power of each band is computed and the difference ineach band from 6 channels gives the spectral powerdifference. These spectral power and spectral powerdifferences are then used by a Multilayer Perceptron (MLPBP)neural network to classify into a baseline state or themental task state. The EEG data were recorded from 4subjects during 2 sessions while the subjects were thinking of4 different mental tasks.
II. METHODOLOGY
The EEG data used in this study were collected by Keirnand Aunon [4]. The subjects were seated in an IndustrialAcoustics Company sound controlled booth with dim lightingand noise-less fan (for ventilation). An Electro-Cap elasticelectrode cap was used to record EEG signals from positionsC3, C4, P3, P4, O1 and O2 (shown in Figure 1), defined bythe 10-20 system [3] of electrode placement. The impedanceof all electrodes were kept below 5 K. Measurements weremade with reference to electrically linked mastoids, A1 andA2. The electrodes were connected through a bank ofamplifiers (Grass7P511), whose band-pass analog filters wereset at 0.1 to 100 Hz. The data were sampled at 250 Hz with aLab Master 12-bit A/D converter mounted on a computer.Before each recording session, the system was calibrated witha known voltage
Download full report
http://csee.essex.ac.uk/Research/BCIs/pu...20band.pdf
